CNN for image Classification

importing libraries

Creating and organizing the images into train, validation and test

Plotting the train sample images

Building and Training a CNN model

Prediction / Inference

We can see that the model is completely overfitting on training data hence the training accuracy is high and the testing accuracy is very low

Building Fine Tuned model using VGG16, a preprocessing technique

Training the fine tuned VGG16 model

Prediction / Inference

We can see that there is a very high accuracy and the model correctly predicted both the classes with an accuracy of 98%

Training custom untrained data by any pre-trained model

MobileNet

Custom dataset (Unseen by imageNet library)

1.png

Modify the Model

Training the model

Predict Sign Language

we get an amazing prediction for the test set with high accuracy

Data Agumentation